FACULTY OF SCIENCE AND ENGINEERING
Linköping Studies in Science and Technology, Dissertation No. 2038, 2020 Department of Physics, Chemistry and Biology
Linköping University SE-581 83 Linköping, Sweden
www.liu.se
Linköping Studies in Science and Technology Dissertation No. 2038 Paw sitiv e selection: Genetics o f dog-human communication 2020
Pawsitive
selection
Genetics of dog-human
communication
Mia Persson
Mia Per ssoni
Linköping Studies in Science and TechnologyDissertations, No. 2038
Pawsitive Selection
Genetics of dog-human communication
Mia Persson
IFM Biology
Department of Physics, Chemistry and Biology Linköpings universitet, SE-581 83 Linköping, Sweden
Pawsitive selection: genetics of dog-human communication © [Mia Persson, 2020]
Cover: Wolf, beagle, golden retriever and Labrador by Linus Nord Andersson
Printed in Sweden by LiU-Tryck, Linköping, Sweden, 2020 ISSN 0345-7524
Abstract
Through domestication and recent selection, dogs have evolved a unique set of communicative skills to attract and redirect human attention. These social skills have not been seen to the same extent in socialised wolves and are therefore believed to have a significant genetic basis. The process of
domestication and breed formation has also had effects on the structure of the dog genome that are favourable for genetic mapping. With a high amount of linkage and long haplotype blocks, fewer genetic markers are needed to find gene-trait associations in dogs than in humans. Dogs serve as an important research model for us since humans and dogs share several diseases, psychiatric disorders and behavioural traits.
In Paper I, I recorded human-directed social behaviours during a two-minute unsolvable problem task in 500 laboratory beagles. The dogs were living at a breeding facility and had been bred, kept and handled under standardised conditions. Behaviours related to task solving and human-directed contact seeking were separated in a principal component analysis, indicating that the behavioural test can be used to study dog-human
interaction. Narrow-sense heritability (h2) of the largest principal component related to contact seeking behaviours was estimated to 0.23. This study found a significant genetic basis to the variation seen in human-directed contact seeking behaviours recorded in this population.
Next, in Paper II, we collected and genotyped the DNA of 190 of the previously tested beagles with an HD Canine SNP-chip. To find genes associated with human-directed contact seeking I performed a genome-wide association study (GWAS), showing one significant and two suggestive single nucleotide polymorphism (SNP) markers on chromosome 26. The significant SNP is located within a gene named SEZ6L, previously associated with autism in human studies. Two adjacent SNPs with suggestive association were found within a gene called ARVCF, which has been associated with schizophrenia. To our knowledge, this was the first genome-wide study to present regions within the dog genome associated with inter-species communication in dogs.
However, these results could have been unique to this beagle population, so Paper III aimed to verify our previous findings in additional dog breeds. We tested 100 Labrador retrievers and 61 golden retrievers with the same unsolvable problem-task used in Paper I. Their DNA was collected and each individual was genotyped by pyrosequencing on two of the previously
identified SNPs. To study the effects of recent selection, the Labrador retrievers were divided into two types. The common type is mainly bred and used for dog shows and as a pet, while the field type is mainly bred and used for hunting purposes. In this study, we found that both markers varied in both dog breeds and was significantly associated with human-directed contact-seeking behaviours. Allele frequencies differed significantly between Labrador
retriever types, suggesting that these loci have been affected by recent selection. In conclusion, Paper III verifies the results found in Paper II.
Finally, in Paper IV we investigated the association between dogs’ human-directed social skills and previously known SNP markers in the
oxytocin receptor (OXTR) region. The oxytocin system plays an important role in the formation of social bonds and may therefore also be important in the bond between dogs and humans. Here, we hypothesized that dogs receiving intranasal oxytocin respond differently to the hormone, depending on the receptor type. To investigate this, 60 golden retrievers were genotyped for SNP markers in the OXTR region and tested with the unsolvable problem task used in Paper I and III. An association was found between genotype and social behaviour in response to oxytocin administration. Dogs responded differently to oxytocin treatment, depending on OXTR genotype. In summary, this thesis contributes to the knowledge on genetic influence of interspecies
Populärvetenskaplig sammanfattning
Hunden har sitt ursprung i vargen och det har gått minst 15 000 år sedan vi människor började domesticera vår bästa vän. Domesticeringsprocessen har lett till att hunden och vargen skiljer sig i en rad olika aspekter.
Utseendemässigt finns det ingen mer variationsrik art än hunden med alla dess raser, storlekar, färger och former. Även beteendet har förändrats under all tid i samlevnad med människan. Om man till exempel ger ett olösligt problem till en varg så kommer den vara envis och försöka gång på gång. Ger du däremot samma problem till en hund så ger den i regel upp snabbt och söker kontakt från människor i närheten, som om den ber om hjälp. Hunden har utvecklat verktyg för att kommunicera med oss på ett sätt som vargen inte gör. Hunden är bra på att söka vår uppmärksamhet men även på att rikta den mot t.ex. en gömd leksak. Även strukturen i hundens genom (DNA) har förändrats, kanske mest på grund av den mer nutida rasaveln. Detta är till fördel för studier som söker efter kopplingar mellan gener och egenskaper så som beteenden, utseendedrag eller sjukdomar. När det kommer till att studera hur gener styr kommunikation mellan två arter så lämpar sig alltså hunden särskilt väl. I artikel I ville vi uppskatta hur stor genetisk effekt vi har på de
kontaktsökande beteenden som hunden visar mot människan. När hundar inte kan lösa ett problem söker de ofta assistans genom framför allt fysisk kontakt och ögonkontakt. Jag har därför använt mig av problemuppgiften för att ge hunden en anledning att söka min uppmärksamhet och sedan har jag mätt hur och hur mycket uppmärksamhet varje hund söker. I den här studien gjorde jag testet på en stor population beaglar på en avelsenhet för försökshundar. Beteenden är ett resultat av samspelet mellan gener och miljön. Vill man studera effekten av det ena är det därför bäst att hålla den andra faktorn så stabil som möjligt. I det här fallet hade alla beaglarna växt upp och levt tillsammans i samma miljö vilket gjorde det lättare för oss att uppskatta hur mycket genetiken bidragit med till den variationen vi ser i beteendet. Med hjälp av stamtavlan kan man räkna ut ett uppskattat värde på beteendets arvbarhet eller hur mycket av skillnaderna i beteendet som går i arv. Hos beaglarna såg vi att nästan en fjärdedel av beteendeskillnaderna kan bero på gener. Vi följde upp den studien med artikel II där vi topsade ca 200 av beaglarna i kinden för att samla DNA. Genom att kolla vilken variant varje individ hade på ca 170 000 markörer i genomet kunde vi hitta samband mellan genetiska markör och beteende. I vår studie fann vi två områden på hundens kromosom 26 som är kopplade till variationen i det kontaktsökande beteendet under problemuppgiften. Samma områden har i andra studier funnits vara kopplade till bland annat autism och schizofreni hos människor.
För att säkerställa att dessa resultat inte bara gäller för just den studerade populationen med beaglar har vi i artikel III även testat golden retrievers och labradorer med samma olösliga problemuppgift. Även hos de raserna fann vi att de två regionerna är kopplade till kontaktsökande beteenden mot
människan. Labradorer avlas dessutom för två olika ändamål, jakt och utställning. Denna indelningen är förhållandevis ny så genom att jämföra frekvensen av markörerna mellan jakt och utställningslabradorer kunde vi även se att den riktade aveln påverkat de genetiska regionerna.
I den fjärde artikeln fokuserade vi på oxytocinreceptor-regionen i hundens genom. Vi vet sedan tidigare att den spelar roll i däggdjurens och alltså även människans och hundens sociala beteenden. Oxytocin är ett må-bra-hormon som utsöndras vid fysisk kontakt och är väldigt viktigt för social anknytning. Det finns olika varianter av DNA-kod för oxytocinreceptorn som hormonet binder till i kroppen. Vår hypotes var att hundar med olika varianter av oxytocinreceptorn kan reagera olika på hormonet vilket kan leda till skillnader i socialt beteende jämtemot människan. För att undersöka detta tog vi DNA-prover från golden retrievers och testade dem med samma olösliga problem som i tidigare studier, med och utan en nässpray med oxytocin. Vi såg att hundarnas kontaktsökande beteenden mot människan ändrades på olika sätt till följd av oxytocinet beroende på vilken variant av oxytocinreceptorgenen hunden hade. Sammanfattningsvis bidrar den här avhandlingen med ny kunskap om hur gener styr hundens unika beteende att söka mänsklig kontakt.
Acknowledgements
This doctoral thesis only has my name on it but as The Beatles sang “I get by with a little help from my friends”. So here are my thank you’s:
First of all, I believe that I have had the best main supervisor possible, Per Jensen. Pelle, thank you for believing in me and for giving me this opportunity. You have given me a lot of trust and responsibility and you have always been there to support.
To my other supervisors, Lina Roth and Dominic Wright. Lina, your support has been invaluable! You always have wise and appreciated input. Your eye for details is incorporated throughout my PhD. And Dominic, thank you for being patient and for guiding me through the winding roads of GWAS. To technicians Ann-Charlotte, Petros and Lejla for the help with equipment and for keeping everything run smoothly in the lab. To my co-authors Ann-Sofie, Martin, Petros, Agaia, Johan and Lise-Lotte for all the work you have put in on my/your/our projects. To Rebecca and Lina for proof reading this wall of text. To all other colleagues at the biology department for contributing to an entertaining work environment and interesting topics for fika-discussions. To my family for always encouraging my academic endeavours. To Mattias for sticking with me and supporting me through all the PhD ups and downs. To our lovely daughters Seia and Idun that has taught me to be really efficient in multitasking and that sleep is overrated.
List of publications included in this thesis
The thesis is based on these listed papers. In the text, they will be referred to as their roman numbers (I-IV).
I. Persson, M.E., Roth, L.S.V., Johnsson, M., Wright, D., Jensen,
P., 2015. Human-directed social behaviour in dogs shows significant heritability. Genes Brain Behav 14, 337-344.
https://doi.org/10.1111/gbb.12194
Re-printed with permission from John Wiley and Sons.
II. Persson, M.E., Wright, D., Roth, L.S., Batakis, P., Jensen, P.,
2016. Genomic Regions Associated With Interspecies
Communication in Dogs Contain Genes Related to Human Social Disorders. Scientific reports 6, 33439.
https://doi.org/10.1038/srep33439
Re-printed under the terms of the Creative Commons CC BY 4.0 license.
III. Persson*, M.E., Sundman*, A.S., Halldén, L.L., Trottier, A.J.,
Jensen, P., 2018. Sociality genes are associated with human-directed social behaviour in golden and Labrador retriever dogs. PeerJ 6, e5889.
https://doi.org/10.7717/peerj.5889
*equal contribution
Re-printed under the terms of the Creative Commons CC BY 4.0 license.
IV. Persson, M.E., Trottier, A.J., Belteky, J., Roth, L.S.V., Jensen,
P., 2017. Intranasal oxytocin and a polymorphism in the oxytocin receptor gene are associated with human-directed social behavior in golden retriever dogs. Horm Behav 95, 85-93.
https://doi.org/10.1016/j.yhbeh.2017.07.016
List of publications not included in this doctoral thesis
Sundman, A.-S., Persson, M.E., Grozelier, A., Halldén, L.-L., Jensen, P., Roth, L.S.V., 2017. Understanding of human referential gestures is not correlated to human-directed social behaviour in Labrador retrievers and German shepherd dogs. Appl Anim Behav Sci.https://doi.org/10.1016/j.applanim.2017.12.017
Jensen, P., Persson, M.E., Wright, D., Johnsson, M., Sundman, A.S., Roth, L.S.V., 2016. The Genetics of How Dogs Became Our Social Allies. Curr Dir Psychol Sci 25, 334-338. https://doi.org/10.1177/0963721416657050
Elfwing, M., Natt, D., Goerlich-Jansson, V.C., Persson, M., Hjelm, J., Jensen, P., 2015. Early Stress Causes Sex-Specific, Life-Long Changes in Behaviour, Levels of Gonadal Hormones, and Gene Expression in Chickens. Plos One 10. https://doi.org/10.1371/journal.pone.0125808
Davies, A.C., Nicol, C.J., Persson, M.E., Radford, A.N., 2014. Behavioural and Physiological Effects of Finely Balanced Decision-Making in Chickens. Plos One 9. https://doi.org/10.1371/journal.pone.0108809
Contents
1 INTRODUCTION ... 1
2 DOMESTICATION AND THE ORIGIN OF THE DOMESTIC DOG ... 2
2.1THE FIRST DOMESTICATED ANIMAL ... 2
2.2ONE ANCESTRAL SPECIES ... 3
2.3BREED FORMATION ... 3
2.4DOMESTICATION AND SOCIAL BEHAVIOUR ... 4
3 DOG-HUMAN COMMUNICATION ... 5
3.1REFERENTIAL SIGNALLING ... 5
3.2DOG-HUMAN EYE CONTACT ... 6
3.3THE UNSOLVABLE PROBLEM PARADIGM ... 6
3.4PERSISTENCE OR SOCIABILITY? ... 8
3.5THE PRESENCE OF THE OWNER ... 9
3.6A GENETIC BASIS ... 9
4 BEHAVIOUR GENETICS ... 11
4.1THE STRUCTURE OF THE DOG GENOME... 11
4.2HERITABILITY OF DOG BEHAVIOUR ... 11
4.3THE DOG AS A MODEL IN GENETIC MAPPING ... 13
4.4GENOME-WIDE ASSOCIATION STUDIES ... 14
4.5CANDIDATE GENE APPROACH ... 16
4.6GENETICS OF HUMAN-DIRECTED SOCIAL BEHAVIOURS ... 16
4.6.1 Heritability of dog social behaviour... 16
4.6.2 Genomic associations with dog social behaviour ... 17
4.7CHALLENGES IN DOG BEHAVIOUR GENETICS ... 18
4.7.1 Environmental factors ... 18
4.7.2 Population stratification ... 18
4.7.3 Long-range haplotypes ... 18
4.7.4 Defining the behaviour ... 19
4.7.5 Epigenetics ... 19
5 OXYTOCIN AND THE OXYTOCIN RECEPTOR GENE ... 20
5.1OXYTOCIN AND THE DOG-HUMAN BOND ... 20
5.2EFFECTS OF OXYTOCIN ADMINISTRATION ... 20
5.3THE OXYTOCIN RECEPTOR GENE ... 21
5.4EFFECTS OF OXYTOCIN AND OXTR GENOTYPE ... 21
6 CONCLUSIONS ... 23
PAPER I ... 24
PAPER II ... 25
PAPER III ... 26
PAPER IV ... 27
1
1 Introduction
The dog was the first animal to be domesticated, the first animal to enter orbit in space and among all other animals valuable to research, the dog was the fourth mammalian genome to be sequenced (Lindblad-Toh et al. 2005). The motive behind the sequencing of the dog genome was the use of the dog as a model for us humans. Dogs share many of our disorders and diseases and the structure of their genome is well suited for genetic mapping. In other words, the dog has become man’s best friend in several contexts.
Through positive selection, artificial selection and just by chance we have created the dog which is, with its almost 400 breeds, probably the most variable vertebrate on earth. However, dogs do not only differ in their morphology but also in behaviour (Mehrkam & Wynne 2014). These behavioural differences and variations are found both within- and across breeds. Combining this behavioural and morphological variation with a genomic structure favourable for genetic mapping research, we could also claim that the dog is the scientist’s best friend.
Dogs have the ability to communicate with another species, us humans (Siniscalchi et al. 2018). They do this willingly, often, in different contexts and to some degree they seem to be born with these skills. The wolf, which is their closest living ancestor, does not possess these human-directed communicative skills to the same extent but can learn to follow our communicative cues by extensive training and socialisation.
The general aim of this thesis was to find genomic regions associated with variation in human-directed social behaviour in dogs tested in an unsolvable problem paradigm. We identified a genetic basis for these behaviours by heritability estimates and found associated genomic regions by a genome-wide association study. These findings were then verified in two additional breeds. Finally, we investigated the effects of oxytocin and oxytocin receptor
2
2 Domestication and the origin of the domestic dog
Animal domestication can be defined as an evolutionary “process whereby populations of animals change genetically and phenotypically in response to the selection pressure associated with a life under human supervision” (Jensen & Wright 2014). This means that domestication is a Darwinian process driven by similar rules, evolving populations through natural selection. However, in the case of domesticated animals, the natural environment is instead created by us humans and the animals have to adapt to this niche that we are providing. In his book, “The variation of animals and plants under
domestication” published in 1868, Charles Darwin uses the variation among dogs to support his theory of evolution through natural selection (Darwin 1868). And even today, the dog is perhaps our foremost example of how much phenotypic variation we can create within a single species by the means of selective breeding.
2.1 The first domesticated animal
The place and timing of dog domestication is still an active debate among scientists of different fields. There are however some general aspects of dog domestication that several studies agree upon. First of all, the dog is both our best and oldest friend, as it was the first animal to be domesticated at least 15 000 years ago (Wang et al. 2013; Wang et al. 2016). Archaeological evidence dating back as far as 14 000 years reveals a close friendly bond between dog and man (Morey 2006). Dogs have not only been buried under ground for hygienic reasons, but they have also been carefully placed in graves together with humans, arranged in sleeping positions, and in some cases, there are even grave offerings present. Archaeological evidence of healed injuries in old dogs also show that they have been treated and cared for during their lifetimes. There are limitations with using archaeological evidence as a means to date the onset of dog domestication. E.g. skeletal findings from 12 000-14 000 years ago highly resemble, or cannot be morphologically distinguished from, wolf remains. Genetic analysis of large canine fossils reveals a possible dog fossil dated to about 32 000 years ago from an archaeological site in Belgium and 15 000 years ago from a location in Russia (Germonpre et al. 2009). Genetic evidence from modern breeds can however tell us more about the timing of a split between wolves and prehistoric dog populations, before significant morphological changes occurred. So far, there does not seem to be a common conclusion upon the exact timing of dog domestication based on these types of genetic analyses as many different timeframes have been suggested, e.g. 15 000-33 000 years ago (Thalmann et al. 2013; Wang et al. 2016) and even >100 000 years ago (Vila et al. 1997). We can conclude that archaeological and genetic evidence combined shows that domestic dogs
3
originated more than 15 000 years ago. This is prior to the development of modern agriculture as well as the domestication of other domestic animals such as the goat, sheep, pig and chicken.2.2 One ancestral species
The second aspect of dog domestication, which the scientific field seems to agree upon, is that no other species than the gray wolf (Canis lupus) has contributed to the dog genome (Ding et al. 2012; Larson et al. 2012; Lindblad-Toh et al. 2005; Wayne 1993). The domestic dog consists of almost 400 different breeds, containing a wide range of morphological variation such as an adult weight between 1 to 100kg, different coat shape and colour or lack thereof, ear shape, leg length and different tail morphologies. It is therefore easy to understand why Darwin believed that dogs were a mix of several species of wild canids (Darwin & Kebler 1859), while all of that variation most likely originates from a very limited number of gray wolves.
2.3 Breed formation
Through genetic research using whole-genome data, it seems as though the Eurasian gray wolf falls out as a sister species to the dog, indicating that this is the closest living dog relative that most likely shares a common wolf ancestor (Fan et al. 2016). Modern breeds, however, are a result of more recent selective breeding as well as admixture between breeds, and have developed over just the past 200 years (Parker et al. 2017; Parker et al. 2004; vonHoldt et al. 2010). There are breeds considered as ancient found in both Asia and Africa that seem to have a shared Asian origin e.g. Shar-Pei, Shiba inu, Akita, Chow Chow and Basenji. The Eurasier falls out in a separate clade close to these ancient breeds, but this is most likely caused by the admixture from several other breeds (Parker et al. 2017). Alaskan Malamute and Siberian Husky also fall out as more ancient breeds but there is evidence pointing towards this being due to more recent wolf admixture (vonHoldt et al. 2010). There are also breeds that were previously thought to be ancient, but genetic analysis shows that they consist of a mixture of modern breeds e.g. the Pharaoh hound (Parker et al. 2004). The Australian dingo (Canis dingo), however, counts as a feral dog and genetic studies on mitochondrial DNA suggest a split from the domestic dog about 5000-8000 years ago (Cairns & Wilton 2016; Savolainen et al. 2004). Unlike the domestic dog, the dingo has not undergone recent
artificial selection and can therefore be an interesting intermediate canine for studying the evolution of behaviours during domestication.
4
2.4 Domestication and social behaviour
Although the details of the initial process are still unknown, prehistoric wolves came so close to co-existing human populations that the humans could take control of their breeding. There are numerous speculations and theories about how the prehistoric wolf became our domestic dog. Many of these are
summarized by Adam Miklosi in the book “Dog behavior, evolution and cognition” (Miklosi 2015). It is believed that the human populations got hold of wolves either by socializing individual cubs taken from the den, or from living very closely to populations of scavenging wolves. Unless humans initially, more or less only, had a ritual relationship with the prehistoric wolves, it is possible that they experienced some advantages by co-existing with them.
My personal point of view is more in line with the “Canine Cooperation Hypothesis” presented by Range & Viranyi (2015). They show that not only dogs, but also wolves, possess a high capability of social attentiveness towards conspecifics. Recently, Range et al. (2019) have also shown that, once overcoming their fear of humans, wolves as well as dogs, are successful when solving a cooperative task together with a human. In theory, together with the fact that the wolf is a highly social and group-living species, this sets a good basis for domestication and the evolution of interspecific cooperation between dogs and humans.
5
3 Dog-human communication
By accepting wolves into our human society, we have created a very unique niche for them to adapt to. You could claim that the natural environment for a dog is living in inter-specific groups with humans, in contrast to conspecifics. This has transformed the social capabilities of wolves into the human-like communicative competence of dogs. Dogs’ abilities to cooperate, communicate and form attachment bonds to humans are all examples of their unique
human-directed social skills (Miklosi & Topal 2013). Sometimes, authors even claim that the relationship between dogs and humans can be called
co-evolution. There are no doubts that we have changed the dog genome through domestication. However, so far there are no convincing evidence of how dogs have changed the human genome. But I do find it interesting that humans seem to have an innate ability to understand dog vocalisations (Farago et al. 2017) as well as emotions just by looking at their faces (Amici et al. 2019) and likewise has been seen in dogs looking at human emotional faces (Siniscalchi et al. 2018).
Another interesting aspect of dog’s social skills is that dogs do not speak “dog” when their communication is directed towards a human. In dog-human interactions, dogs still use the same signals as when communicating with conspecifics, but with another meaning. E.g. dogs readily use eye-contact when interacting with humans and this facilitates the dog-human bond while eye-contact in dog-dog interactions are perceived as threatening (Siniscalchi et al. 2018).
3.1 Referential signalling
Referential gestures, such as pointing or gazing, are produced in order to direct the attention of a recipient towards e.g. an object. Human infants use referential gestures as a prelinguistic means of communication (Liszkowski et al. 2012), but also dogs can communicate with us in a similar manner. Dogs are able to find hidden objects by the use of indirect human cues and they can also communicate this information to their owners (Lakatos et al. 2012). Worsley & O'Hara (2018) describe 19 different intentional referential gestures used by dogs in everyday communication events with their owners. Behaviours directed towards a human such as eye contact, different types of physical contact as well as vocalisations are examples of how dogs seem to re-direct our attention towards e.g. an object. One of the most studied types of referential signalling or “showing” behaviour in dogs is gaze alternation between the recipient and a desired out-of-reach object (Miklosi et al. 2000; Worsley & O'Hara 2018). If an object is hidden in the absence of the owner, the dog will
6
display gaze alternation even more frequently, suggesting an intention to inform their owner of the hidden object (Viranyi et al. 2006). Vocalisations can also occur together with gaze alternations as a means to direct the owners’ attention towards the hiding location (Miklosi et al. 2000).
3.2 Dog-human eye contact
Eye-contact is a key stone in the communication and social attachment bond between dogs and humans. Several studies have reported significant differences between dogs and wolves in human-directed gazing behaviour during problem solving (e.g. Miklosi et al. (2003)). This difference is also evident when
comparing wolves to pet dogs as well as shelter dogs (Udell 2015). Encouraging behaviour from the owner prior to problem solving can increase gazing
behaviour (Horn et al. 2012). Also, age and/or experience have an effect on eye-contact seeking behaviour where older dogs seem to utilise this behaviour more frequently as seen in paper I and Passalacqua et al. (2011). Passalacqua et al. (2011) also saw that gazing behaviour can vary between breeds as they found it to be more frequent in hunting and herding dogs. Jakovcevic et al. (2010) saw that retrievers gazed more towards the human face than German shepherds and poodles. Administration of the neuropeptide oxytocin increases gaze towards the eye region in humans (Graustella & MacLeod 2012) and dogs (Nagasawa et al. 2015; Nagasawa et al. 2017). Interestingly, eye-contact itself can increase oxytocin levels in both the dog and its owner but this is not found between wolves and their caretakers (Nagasawa et al. 2015). The oxytocin system, which is explained further in section 5, could explain the importance of eye-contact on the dog-human attachment bond.
3.3 The unsolvable problem paradigm
One situation where dogs typically display referential gestures towards a human, such as eye contact or physical contact, is when they are faced with a problem-solving situation. Dogs increase their human-directed gazing when a task becomes unsolvable but also depending on the attentional stance of the recipient (Marshall-Pescini et al. 2013). This suggests that gaze alternation during an unsolvable-task situation is both an intentional as well as a
referential behaviour. The unsolvable task paradigm can be used to provoke these referential gestures in dogs in order to study them systematically as we have done in paper I, III & IV. An unsolvable problem-task can either be a solvable task that has been modified to become unsolvable in the actual testing situation or the task can be unsolvable from the start. If dogs are tested with the first alternative, they will all need prior training to learn how to solve the task. In paper I, I tested 500 beagles with an unsolvable problem-solving task. To also train all of these individuals prior to testing would just have been too
7
time consuming. Instead, Dr Lina Roth and Prof Per Jensen at Linköping University designed a test apparatus that is both solvable and unsolvable at the same time (Figure 1). It consists of three identical problems where dogs can access treats hidden underneath lids. Two of these lids can be easily pushed open while the third lid cannot be opened. With this apparatus, the dogs quickly learn how to access the treats by trial and error while the final problem is unsolvable.Figure 1. The unsolvable task used in paper I, III and IV. Treats are hidden in
the circular compartments underneath the Plexiglas lids. The lids to the left and right can be opened while the middle compartment remains inaccessible to the dog.
8
3.4 Persistence or sociability?
Topal et al. (1997) have shown that the relationship between dogs and their owners affects their problem-solving abilities. They saw that companion dogs were more socially dependent on the owners and did not perform as well in a problem-solving situation as dogs with a working-related relationship with their owners. This could be interpreted such that the decrease in problem solving ability is not a product of domestication, but rather the strong social
attachment between the dog and its owner. However, Udell (2015) did not find a difference between pet and shelter dogs in problem solving abilities with a neutral person present.
On the other hand, it has been argued that differences measured between dogs and wolves, in human-directed social behaviours in a problem-solving
situation, is a result of differences in persistence rather than sociability (Brubaker et al. 2017; Marshall-Pescini et al. 2017; Rao et al. 2018; Udell 2015). Studies have found a negative correlation between time spent
manipulating the task and eye-contact (Brubaker et al. 2017; Marshall-Pescini et al. 2017). This could be interpreted such that individuals that are more persistent or task motivated do not have the time to also seek human attention. Rao et al. (2018) showed that wolves are more persistent or task oriented than dogs, even in a situation where a human is not present. This could show that wolves are more motivated than dogs to manipulate objects. Often in problem-solving tasks, there is a food reward involved (e.g. Miklosi et al. (2003); Udell (2015)) and so is the case also in our unsolvable problem setup in paper I, III & IV. In these types of setups, it is possible that differences in task
manipulation correlates to differences in food motivation. One way of attempting to account for food motivation, is by testing animals that are at a similar level of hunger. In paper I, III & IV we performed a brief food motivation pre-test where dogs not willing to eat the treats were excluded from testing. In a test-setup involving different species, such as dogs and wolves, however, the problem of food preference also needs to be addressed. In paper I we summarised behaviours displayed by beagles during an
unsolvable task situation in a principal component analysis (PCA). This showed that behaviours directed towards the task did not seem to share variance with human-directed behaviours. Similar findings were reported by Sundman et al. (2017). However, in this case we should keep in mind that genetic markers were found to be associated with the actual behavioural measurements and not with the principal components (paper II).
One way of interpreting these joint findings is that we should be careful in directly comparing dogs’ and wolves’ human-directed and task-oriented
9
behaviours evaluated with the unsolvable problem paradigm. Perhaps this is also something to take into consideration when testing different dog breeds.3.5 The presence of the owner
Another factor to keep in mind is the presence or absence of the owner. There are studies showing that the owner can act as a secure base for the dog (Mariti et al. 2013; Topal et al. 1998). In those cases, the presence of the owner increased both explorative and play behaviours during testing. In paper I we tested beagles bred at a kennel for laboratory dogs. These individuals did not have owners, but a set of caregivers. During testing there was only one human (me) present in the room and she was unfamiliar to them. In paper III & IV on the other hand, we tested pet dogs and, in these studies, there were always one owner and an unfamiliar experimenter present in the arena during testing. This setup allows us to record behaviours directed towards both a familiar and an unfamiliar human during the problem-solving situation.
3.6 A genetic basis
Dogs form an attachment bond to their owners similar to the bond between a child and the parent (Siniscalchi et al. 2013; Topal et al. 1998). The strong attachment bond between dogs and their owners is evident already at an early age. However, even extensively socialised wolves do not form this type of relationship with humans (Topal et al. 2005). Miklosi et al. (2003) identified what seems to be a fundamental difference between dogs and wolves using the unsolvable task paradigm. When faced with an unsolvable problem, dog give up and turn to a nearby human in a “help-seeking” manner while wolves do not. Furthermore, dogs, but not wolves, use mutual gazing as a mean of communication with humans (Nagasawa et al. 2015). Generally, even fully socialised wolves do not seek as much human attention as dogs do (Gacsi et al. 2009). This is evident already at an early age (Gacsi et al. 2005). Dogs seek human eye-contact more often and for a longer duration than wolves
(Nagasawa et al. 2015). The dingo seeks less human eye-contact than dogs, but more than wolves (Johnston et al. 2017). This is interesting as the split between dogs and dingoes occurred at least 5000 years ago (Figure 2) and therefore, the dingo has undergone early domestication but not recent selection (Cairns et al. 2017). This suggests that genetic changes developing during the course of domestication may have affected dogs’ sociality. Thus, genetics of interspecies communication is a central aspect of domestication.
10
Figure 2. Left: a timeline displaying the historical split between wolves and dogs approximately
15 000 years ago. The split between semi-domesticated dogs and dingoes occurred
approximately 5 000 years ago, prior to modern dog breed formation. Therefore, the dingo has undergone early domestication but not recent artificial selection. Right: kernel density plot adapted from Johnston et al. (2017), showing the probability density function of log-transformed eye-contact duration of wolves, dingoes and dogs.
Figure 3. A demonstration of haplotype pattern across and within dog breeds. Intense selection
and inbreeding during modern dog breed-creation has affected the haplotype structure of their genome. Within breed, haplotype blocks are long-ranged and breed-specific. Across breed, the underlying short-range ancestral haplotypes are still retained.
11
4 Behaviour genetics
Behaviour genetics is not the study of how genes affect behaviour, but rather the study of how genes and the environment affect behaviour. Most behaviours are complex traits that vary depending on both genetic and environmental factors. We could use breed specific hunting or herding behaviours as an example. These behaviours have clearly been selected for in the specific breed and therefore have a genetic basis to them. However, behaviours such as herding can be further shaped by training, which is an environmental factor. To make this even more complex, we could also study how the interaction between genes and environment (epigenetics) affect behaviour, but that is outside of the scope of this thesis.
4.1 The structure of the dog genome
At least two population bottlenecks have had significant effects on the structure of the dog genome (Lindblad-Toh et al. 2005). The first bottleneck was rather mild and is related to the initial domestication from wolves at least 15 000 years ago. The second is a result of the intense artificial selection that has occurred during recent breed formation, creating nearly 400 different dog breeds in the last 200 years. A small subset of individuals has founded each breed, and as a result, their specific long-range haplotype patterns are common within that breed and has not yet been separated by recombination (Figure 3). These long-range blocks of haplotypes are inherited together more often than at random, and this is called that they are in linkage disequilibrium (LD). These linked blocks are on average in megabases within dog breeds while human and across dog breed LD extends on average in kilobases (Hall & Wynne 2012; Lindblad-Toh et al. 2005; Wall & Pritchard 2003).
4.2 Heritability of dog behaviour
Scott & Fuller (1965) conducted over a decade long selective breeding experiments in five dog breeds, revealing the pattern of inheritance and variation in numerous behaviours. Fortunately, dog breeds act as genetically isolated populations selected for specific morphological and behavioural traits. Thus, the genetic influence on dog behaviour can be studied, without creating your own selective breeding experiment, by estimating the heritability of a trait using pedigree information, as in paper I. A trait can include any kind of characteristics that varies within the population such as disease, morphological and behavioural differences. The heritability of a trait explains how much of the variation measured in a population that can be accounted to genetic
12
factors (Visscher et al. 2008). The estimate of heritability is specific to both population and environment and can change over time.
Each behaviour displayed is a product of both environmental and genetic factors. Variation in the phenotype (VP) can be explained as the sum of the underlying genetics (VG) and environmental variance (VE) (Wilson et al. 2011). Broad-sense heritability (H2) is defined as the proportion of phenotypic variance, in e.g. a behaviour, that can be explained by environmental or genetic factors as shown below.
𝐻"=𝑉%
𝑉&
However, VG includes all genetic factors that can possibly affect the behaviour such as dominance (VD), epistasis (between loci allele interactions, VI) and additive effects (average allele effects, VA) (Wilson et al. 2011). On the other hand, VD and VI are difficult to estimate, and since only one allele of each parent is inherited at a given locus, VA is the variable mostly explaining parent-offspring resemblance. Therefore, it is most often narrow-sense heritability (h2) that is estimated in an animal model. Narrow-sense heritability is defined as the proportion of phenotypic variance that can be explained by environmental or additive genetic effects.
ℎ"=𝑉(
𝑉&
Heritability estimates range from 0 to 1. If the heritability of a trait is estimated to 0, all of the variability of the trait is due to environmental factors. Political preferences or religion are examples of characteristics with a heritability estimate close to 0 among human populations. On the contrary, a heritability estimate of 1 would mean that the phenotypic variation is completely
dependent on genetic variation. Disorders caused by single-gene mutations are examples of traits with a heritability close to 1. However, behavioural traits typically have an intermediate h2 since they, like most traits, are complex and their variability is due to a combination of environmental and genetic factors. In paper I we estimated h2 of a principal component related to human-directed social behaviours in a population of laboratory beagles to 0.23. This means that 23% of the variability in human-directed social behaviours in this particular population could be accounted to genetic differences among the dogs. This was estimated based on the entire pedigree information available including 643 individuals in total and 437 of these were tested. In paper II however, we calculated the so called “chip heritability” which is based on a
13
genetic relationship matrix of the 190 genotyped individuals. The genomic relationship matrix uses the genotyped single-nucleotide polymorphisms (SNPs) (genetic markers) to estimate genomic share among the individuals. We can then statistically model the proportion of trait variance explained by this genomic sharing. Our chip heritability of the same principal component of human-directed social behaviour from paper I was 0.14 in contrast to 0.23. The chip heritability of the direct behavioural traits varied from 0.37 for duration spent in human proximity to 0.05 for frequency of visiting human proximity. You can find all the chip heritability estimates of the behavioural variables in Table 1 of paper II.Heritability estimates do not provide any information about which or how many environmental factors or genes that are involved. However, it can indicate whether there is sufficient genetic contribution to a trait of interest, such that a genome-wide association study (GWAS) could reveal associated loci.
4.3 The dog as a model in genetic mapping
There are numerous animals valuable for research, but the dog was still the fourth mammalian genome to be sequenced (Lindblad-Toh et al. 2005). The reason for this was argued on the use of dogs as models for human disease and the structure of the dog genome makes them particularly suitable for
association mapping research (Lindblad-Toh et al. 2005; Sutter et al. 2004). Because of these long-range LD blocks within breeds, fewer genetic markers are needed in genome-wide association studies (GWAS) (Lindblad-Toh et al. 2005; Wall & Pritchard 2003). In humans, where LD blocks are approximately 50 times shorter, at least 300 000 markers are required while 10 000-30 000 should be sufficient in dogs. Thus, it is more economical to perform GWAS in dogs than in humans. In paper II, we used a high-density canine SNP chip with more than 172 000 SNP markers to search for behaviour associations. The great variation in both morphology and behaviour among dogs makes them a gold mine for genetic mapping. Morphologically, dogs are generally very similar within breeds but on the contrary, very different between breeds. Dog breeds differ greatly in e.g. size, fur (colour, texture, length), skull shape, tail shape (Shearin & Ostrander 2010), but also in specific behaviours such as pointing and herding. Variation in dog behaviour is similar to what is observed in human populations. E.g. dogs show differences in personality and
temperament (Svartberg & Forkman 2002; Svartberg et al. 2005), for review see (Jones & Gosling 2005; Mehrkam & Wynne 2014), aggression (Svartberg 2006; van den Berg et al. 2003) and anxiety (Tiira et al. 2016).
14
No other species on earth communicate with us humans like dogs do. Therefore, dogs serve as a unique model for us to study the evolution and genetics of interspecies communication. But dogs are also an important genetic model for us to learn more about human diseases and disorders. We do not only share our homes and similar behavioural variation with dogs, but also welfare diseases such as obesity (German 2006) and diabetes (Niaz et al. 2018) as well as psychiatric disorders such as obsessive compulsory disorder (Moon-Fanelli et al. 2011) and hyperactivity (Kubinyi et al. 2012). Up until today, 746 diseases or disorders have been reported in dogs and 438 of these are potential models for human disease (OMIA, Online Mendelian Inheritance In animals,
https://omia.org/home/). Interestingly, the regions we identified to be associated with human-directed social behaviours in the beagle population (paper II) contain genes previously associated with e.g. autism (Chapman et al. 2015), bipolar disorder (Xu et al. 2013), ADHD (van Goozen et al. 2015), aggression (Qayyum et al. 2015) and schizophrenia (Chen et al. 2005) in humans.
4.4 Genome-wide association studies
If you have found variation in a trait that you suspect has a significant genetic basis, but you do not have any candidate genes, a genome-wide association study (GWAS) can be used to find candidate regions.
SNPs are DNA sequence variations, commonly used as genetic markers, where one nucleotide differs between members of a species. At one given locus there might be two alleles present in a population, let us say A (adenine) or C (cytosine). These can also differ within the chromosome pair of one individual making it heterozygote, or homozygote if it has the same variant on both chromosomes. So, if the SNP is available as an A or a C, you can find
individuals that are either AA, AC or CC at this locus within your population. GWAS is a high-throughput method of genotyping SNPs throughout the entire genome (Anholt & Mackay 2010). This is a powerful method in case-control studies where genotype frequencies are compared between case and control individuals to search for statistical associations between sequence variants and the studied disease. However, GWAS can also be used to find associations between genetic markers and a quantitative trait, e.g. behaviours. In paper II we used a high-density canine SNP-chip to genotype almost 173 000 SNPs evenly spaced throughout the entire genome in 190 laboratory beagles. The within-breed approach has the benefit of reducing locus
heterogeneity, similarly to the advantages of using isolated human populations in countries like Finland or Iceland (Ostrander & Kruglyak 2000). In this analysis, we identified two regions on chromosome 26 associated with
human-15
directed social behaviours. One marker located in an intron (non-coding region) of the SEZ6L-gene was significantly associated with the duration spent in human proximity. SEZ6L was the only gene present within its linkage block (Figure 4). We also found two adjacent markers within introns of the ARVCF-gene, suggestively associated with the duration spent in physical contact. Within the same linkage block of the ARVCF-gene there are three additional genes: TXNRD2, COMT and TANGO2. Out of these genes, the COMT-gene is highly interesting as it has previously been associated with social disorders and schizophrenia (Sanders et al. 2005).One of the limitations of the GWAS is that it only identifies statistical
associations between the trait and the genetic markers. It does not necessarily give you information about which gene within a linkage block that is involved but it rather gives you a candidate region for further analysis such as fine-mapping or a candidate gene approach. This has been nicely done by
vonHoldt et al. (2010) who identified a genomic region under positive selection in dog breeds, located on chromosome 6, by an across-breed genome-wide approach. A deletion of this region in humans is known to cause William-Beuren syndrome where one of the characteristics is hyper social behaviour. Later, they identified structural variants, in two genes (GTF2I and GTF2IRD1) contained in this region, associated with extreme sociability in dogs (vonHoldt et al. 2017). Another example is Dodman et al. (2010) who performed a GWAS to identify a candidate region associated with canine compulsive disorder (CCD). By fine-mapping, the gene CDH2 was revealed to be involved in compulsive behaviour in Dobermann pinchers.
Figure 4. The linkage blocks on chromosome 26 associated with proximity and physical
16
4.5 Candidate gene approach
A candidate gene approach can be used within behavioural genetic research if there is a suspected gene or region of interest. Instead of genotyping thousands of genetic markers across the genome, the focus could be on just one or a few within the candidate region. In our case, the behavioural variation and heritability estimates from paper I served as a basis for the genome-wide analysis in paper II resulting in two candidate regions associated with human-directed social behaviour in dogs. However, these results could be unique to this specific laboratory beagle population. Therefore, we decided to use a candidate gene approach in order to verify the marker-behaviour association, found in paper II, in two additional breeds (paper III). The candidate gene approach was also used in paper IV where previously identified SNPs within the oxytocin receptor gene (OXTR, see section 5) were genotyped in golden retriever dogs. In paper III & IV the SNP genotypes were determined by pyrosequencing (Ahmadian et al. 2006).
4.6 Genetics of human-directed social behaviours
We have shown that human-directed social behaviour varies within breeds (paper I & III) and there are studies showing that they differ between breeds (Passalacqua et al. 2011; Sundman et al. 2018; Udell et al. 2014). Interestingly, these types of behaviours have also been shown to differ between breed types recently diverged by artificial selection (Sundman et al. 2016). One example of this is the Labrador retriever that has been bred into two selection lines based on their purpose as pet and show dogs or as hunting dogs. These recent selection lines give us a unique opportunity to study the effects of very recent artificial selection on the evolution of behavioural traits as well as changes in allele frequencies as utilised in paper III.
4.6.1 Heritability of dog social behaviour
As mentioned earlier in section 4.2, heritability estimates can give us a clue about how much of the variation in a behaviour measured in a population that is due to additive genetic factors. Heritability (h2) has been estimated for different types of human directed social behaviours in dogs. Liinamo et al. (2007) estimated the heritability of human-directed aggression in golden retrievers to be 0.77, which can be considered as a very high h2 estimate for a behavioural trait. Sundman et al. (2016) estimated the heritability of PCA components from behaviours evaluated during a Swedish temperament test called Dog Mentality Assessment (DMA). A study in Rough Collie, estimated heritability of behavioural traits evaluated with DMA, as well as a
17
of the DMA personality trait called sociability (0.22) and C-BARQ behavioural subscale of human-directed play interest (0.28), attachment/attention seeking (0.10) and stranger-directed interest (0.29) to mention a few. Furthermore Wilsson & Sundgren (1998) estimated the heritability of social contact in 8-week-old German shepherd puppies to 0.42. Van der Waaij et al. (2008) estimated the heritability of dogs’ willingness to seek human contact (subjective score) to 0.38 in German shepherds and 0.03 in Labrador retrievers. In paperI, we estimated h2 of human-directed social behaviours to 0.23.
4.6.2 Genomic associations with dog social behaviour
In paper II, a genetic marker on chromosome 26 within the SEZ6L-gene was found to be associated with the duration of physical contact seeking and the duration spent in human proximity in beagles. The same marker was associated with physical contact seeking in golden retrievers and duration of owner gazing within field type Labradors (paper III). The genetic marker within the ARVCF-gene, also on chromosome 26, was suggestively associated with the duration beagles spent in physical contact with a human in paper II. In paper III this marker was associated with gaze frequency, human
proximity and physical contact seeking in golden retrievers. Additionally, in Labrador retrievers the ARVCF-marker was associated with human proximity and gaze frequency. Comparing the gaze behaviour in Labrador retriever selection lines, only the field type differed depending on marker genotype. Also, genotype frequencies of these markers differed between the selection lines, suggesting that there is a genetic basis underlying this behaviour that could have been selected for recently.
To my knowledge, no other studies have found associations between dog behaviour and genetic markers in the regions of the SEZ6L or ARVCF genes. However, recently, Caniglia et al. (2018) described the genetic architecture of a dog x wolf crossbreed and found regions containing excess of dog and wolf genes respectively. The Czechoslovakian Wolfdog is a cross between German shepherd dogs and Carpathian wolves that look like a wolf but behave more like a dog. Interestingly, both genomic regions we suggested to be associated with human-directed contact seeking, based on the GWAS in beagles (paper II), are contained within regions of excess dog ancestry.
In addition to the two genes suggested in paper II, there are also other genes suggested to be associated with human-directed social behaviour in dogs. One of these is the oxytocin receptor gene (OXTR) (Kis et al., 2014; paper IV) that will be discussed in more detail further on in section 5. Another example is the dopamine- and serotonin-related genes associated with human-directed aggression (Våge et al. 2010). Furthermore, I have previously mentioned the
18
genomic region related to William-Beuren syndrome in humans and hypersocial behaviour in dogs in section 4.4 (vonHoldt et al. 2017).
4.7 Challenges in dog behaviour genetics
Although there are several advantages of using canines as a model animal for genetic mapping, there are also challenges.
4.7.1 Environmental factors
There has been great success in mapping many mendelian (single gene) traits in dogs, but the advancement in the mapping of complex traits is still
struggling. Behaviours are typically complex traits affected by multiple small genetic effects and environmental factors such as experience and learning (Hall & Wynne 2012; van Rooy et al. 2014). It is much easier to entangle underlying genetic factors if the environmental effects are kept at a minimum. The
experience of communicating with humans increases throughout the dogs’ lifetime so one way of decreasing the effect of this could be to only include individuals of similar age. For paper I and II we had the great advantage of working with a population of beagles at a breeding facility for laboratory dogs. Hence, the majority of individuals were young (approximately 1,3 years). Additionally, as they were bred, kept and handled at the same facility, environmental effects were at a minimum.
4.7.2 Population stratification
On the other hand, an isolated population comes with the problem of
population structure or population stratification often due to varying degrees of inbreeding. Population stratification can e.g. lead to false positive results and has to be adjusted for in a GWAS. In our case, population stratification ranged from 1.0-1.4 which is suggestive of minor population stratification in our specific beagle population (paper II). To adjust for both population stratification and relatedness we performed the GWAS using the GEMMA (Genome-wide Efficient, Mixed Models Association) software (Zhou &
Stephens 2012). Another advantage of the mixed models analysis for GWAS is that you can include fixed effects in your model. In our case we included effects of sex and age.
4.7.3 Long-range haplotypes
I have previously mentioned the long haplotype blocks and linkage disequilibrium in the dog genome as an advantage in genetic mapping research. However, once you find significant marker associations, these may
19
occur in large linkage blocks containing numerous genes, making it a difficult task to determine the causal variants.4.7.4 Defining the behaviour
Initially, it is crucial that the behavioural phenotype is carefully defined when searching for gene associations. In dog behaviour research, it is common to use test batteries and questionnaires to measure behavioural traits. The behaviours are then summarised or grouped with either factor analysis (FA) or principal component analysis (PCA) that identifies correlated variables. E.g. in paper I we measured behaviours in beagles during a problem-solving task. Using a PCA we could see that behaviours directed towards the problem apparatus loaded in a separate component to human-directed social behaviours,
indicating separate correlational factors. Just because traits factor together does not necessarily mean that they share a common genetic foundation. An FA or PCA cannot distinguish shared genetic and environmental factors. One way of dealing with this is by deconstruction of the factor-based phenotypes into their underlying components. This is what we did in paper II when we searched for behaviour-gene associations of the actual behaviours contained within the human-directed social behaviour component from the previous PCA.
4.7.5 Epigenetics
Although there are great morphological and behavioural differences between dogs and wolves, they are genetically very similar (Kirkness et al. 2003). This indicates that many of these differences, that have evolved very rapidly by recent selection, are not due to differences in gene structure but rather in gene expression. In that case, these genetic effects will not be picked up by a GWAS approach.
20
5 Oxytocin and the oxytocin receptor gene
We performed a GWAS in order to identify novel genomic regions associated with human-directed social behaviour in dogs. However, based on research done in other mammals, and humans, there is a previously identified candidate gene with potential association to dogs’ unique social skills. This gene is called the oxytocin receptor gene (OXTR-gene) and it is well conserved in mammals (Gimpl & Fahrenholz 2001).
5.1 Oxytocin and the dog-human bond
The neuropeptide oxytocin has a well-established function in social bonding in humans and other mammals (Lim & Young 2006). In humans, there are several documented psychological and physiological health benefits of human-animal interaction and the oxytocin system is believed to play a key role (Beetz et al. 2012). These effects seem to be mutual since dog-human social
interactions increase blood oxytocin levels and decreases heart rate in both dogs and humans (Handlin et al. 2011; Odendaal & Meintjes 2003). It has been suggested that oxytocin also plays a fundamental role in forming the social bond between dogs and humans. Nagasawa et al. (2015) found that eye-contact from dogs, but not wolves, increased urinary oxytocin in humans. In turn, this increased humans’ affiliative behaviours towards the dogs causing an increase in dogs’ urinary oxytocin levels. They also showed that eye-contact increased following oxytocin administration. This proposes a gaze-modulated positive feedback-loop as a modulator for the evolution of the dog-human bond. However, recently, Marshall-Pescini et al. (2019) did not find any effects of dog-owner interaction on urinary oxytocin levels in neither the dogs nor the owners.
5.2 Effects of oxytocin administration
Oxytocin can act as both a neurotransmitter and a neurohormone and is produced in the periphery as well as in the hypothalamus (Gimpl &
Fahrenholz 2001; Neumann 2008). It has been proposed that behaviours are mainly affected by central oxytocin as opposed to peripheral oxytocin (Leng & Ludwig 2016). Peripheral and central oxytocin can be experimentally
manipulated by intranasal administration of the hormone, although the mechanisms seem yet to be poorly understood and the increase in brain oxytocin is minor (Neumann et al. 2013; Rault 2016). Even so, intranasally administrated oxytocin has been shown to modulate dogs processing of human emotional faces (Kis et al. 2017), increase eye-contact with the owner
21
(Kis et al. 2015), increase affiliative behaviours towards owners andconspecifics (Romero et al. 2014) and modulates dogs response to inequity (Romero et al. 2019). However, the effects of oxytocin are not always pro-social and can be somewhat unpredictable depending on the individual and context (Bartz et al. 2011). Oxytocin administration can decrease cooperation and trust towards strangers or outgroups in humans (De Dreu 2012; De Dreu et al. 2010; De Dreu et al. 2011) and decrease friendliness in dogs approached by a threatening human (Hernadi et al. 2015). Additionally, breed differences have been described in effects of oxytocin administration (Kovacs et al. 2016).
5.3 The oxytocin receptor gene
One explanation to these contradictory results could be that individuals respond to oxytocin differently depending on OXTR genotype. In humans, variation in the OXTR gene has been associated with empathy and stress reactivity (Rodrigues et al. 2009), autism (Jacob et al. 2007), temperament (Tost et al. 2010) and attachment style (Denes 2015). Also, in humans, OXTR genotypes have been associated with different behavioural responses to
oxytocin administration (Feng et al. 2015; Marsh et al. 2012). The OXTR gene is not as well studied in dogs. However, Kis et al. (2014) found associations between polymorphisms in close proximity of the OXTR gene and social behaviours, such as friendliness and proximity seeking in German shepherds and border collies. Interestingly, the opposite relationship between genotype and behaviour was seen between breeds, highlighting the importance of taking breed into account when studying the effects of oxytocin. Also, Turcsan et al. (2017) studied the association between four OXTR polymorphisms and social affiliative (greeting) behaviour in border collies. They found that the
polymorphisms were associated with greeting behaviour but was context and in some cases sex dependent. The authors suggest that the genetic markers in the OXTR region might be related to different functions in the oxytocin system.
5.4 Effects of oxytocin and
OXTR
genotype
Considering the behavioural effects of oxytocin on dogs, genetic variation in the oxytocin system may affect the bond between dogs and owners as well as between- and within-breed differences in human-directed social behaviour. In paper IV, we wanted to investigate associations between known
polymorphisms in close proximity of the OXTR-gene (Figure 5), and the effects of intranasal oxytocin treatment on human-directed social behaviours in dogs. To do so, golden retriever dogs were subjected to the same behavioural test as in paper I and III, the unsolvable task. We found that oxytocin treatment decreased physical contact seeking from the unfamiliar experimenter and this was the only significant effect found of oxytocin administration on a group
22
level. Independent of oxytocin treatment, a significant association was found between one of the OXTR-SNP genotypes and latency to seek physical contact from the owner. However, taking both oxytocin treatment and OXTR
genotype into consideration, we found that individuals with the AA genotype tended to increase while GG individuals significantly decreased their frequency of seeking owner physical contact after oxytocin administration. These results indicate that the contradictory effects, or lack of effects, of oxytocin
administration could be due to genetic variation in the oxytocin receptor.
Figure 5. A schematic figure of the OXTR-gene on canine chromosome 20 starting with a short
un-translated region (UTR). The SNP-markers studied in paper IV are positioned just before the first and after the last exon and are marked in purple.
23
6 Conclusions
Dog sociability is likely a heritage from their wolf ancestors that has been shaped by humans trough thousands of years of domestication and selective breeding. Human-directed social behaviours in dogs are complex traits, in which variation mostly depends on environmental factors such as experience. By using the unsolvable problem paradigm, social behaviours directed towards humans can be empirically and systematically studied. In paper I we show that there is a significant genetic basis to the variation in social behaviours directed towards a human in a population of laboratory beagles. We then continued with a genome-wide association study that revealed two candidate regions on chromosome 26 associated with this variation (paper II). These genomic regions associated with dog social behaviour, are associated with social disorders in human such as autism and bipolar disorder. To ensure that our findings were not limited to the specific beagle population, we also investigated the genotype-phenotype association in two additional breeds, golden and Labrador retrievers (paper III). Indeed, a similar association between genotype and behaviour was found in both of these breeds as well. In addition, the difference in genotype frequencies between the two recently separated Labrador retriever types indicates that these genomic regions could have been affected by selection. Finally, we show that some of the variation in dogs’ social behaviours directed towards humans might be accounted to variation in the oxytocin receptor gene (paper IV).
24
Paper summaries
Paper I
Human-directed social behaviour in dogs shows significant heritability.
At least 15 000 years of domestication and selective breeding has changed the ancestral wolf into the modern dog. Dogs seem to have evolved unique social skills enabling them to communicate and cooperate with humans. They are able to seek our attention as well as re-directing it towards e.g. an out of reach object. However, there is individual variation in attention-seeking behaviour. In paper I, we wanted to investigate wether we could estimate the genetic basis of the variation in human-directed contact seeking behaviour. To do this, we tested 500 laboratory beagles in an unsolvable problem paradigm. Task- and human-directed behaviours were summarised in a principal component analysis, showing that they appear to have separate underlying mechanisms. We found that sex and age affect human-directed contact seeking and the heritability of human-directed social behaviours were estimated to 0.23. This indicates a significant genetic basis to the behavioural variation found in this population.
Figure 6. Effects of sex on
average principal component scores. Error bars: ± SEM
25
Paper II
Genomic regions associated with interspecies communication in dogs contain genes related to human social disorders.
In paper II we wanted to follow up on the results from paper I. We had previously identified a significant genetic basis, to the variation measured in human-directed social behaviours, in a large population of laboratory beagles. Now, our aim was to identify novel genomic regions associated with these human-directed behaviours. DNA was collected from the 95 most and 95 least social individuals based on their principal component scores from paper I. These individuals were then genotyped with an HD Canine SNP chip containing almost 173 000 SNP markers, evenly spread throughout the genome. Through a genome-wide association study (GWAS) we found two regions on chromosome 26 significantly and suggestively associated with the duration of human proximity and physical contact seeking. Interestingly, these regions contain genes previously associated with social disorders such as autism, bipolar disorder and agression in humans.
Figure 7. Manhattan plot from the genome-wide association analysis on the duration spent in
physical contact. One marker within the SEZ6L-gene on chromosome 26 reaches significance (red line, Bonferroni 5%). Two adjacent markers within the ARVCF-gene, also on chromosome 26, reaches suggestive p-value (pink line, Bonferroni 10%).
26
Paper III
Sociality genes are associated with human-directed social behaviour in golden and Labrador retriever dogs.
One of the draw backs of paper I and II is that they are based on one isolated population of laboratory beagles. This means that the association we have identified between genetic markers and human-directed social behaviour are unique to this population. Therefore, in paper III, we aimed to verify our findings by testing two additional breeds, golden and Labrador retriever, using the unsolvable task paradigm. Also, the Labrador retrievers have been split into two separate selection lines for approximately 70 years. One line of breeding is focused on producing pet dogs for confirmation and show purposes while the other line is selected based on their hunting qualities. This gave us the opportunity to also compare allele frequencies of the genetic markers between the selection lines. In this study, we found significant associations between the genetic markers and human attention-seeking in both breeds. We also found differing allele frequencies between the Labrador selection lines. These results verify the findings in paper I and also suggest that these genomic loci may have been affected by recent selection within Labrador retrievers.
Figure 8. A) Association between SNP1-genotypes (the SEZ6L-marker) and the duration of
owner-directed gazing in Labrador retrievers (red/pink) divided by type. B) Associations between SNP2-genotypes (the ARVCF-marker) and the frequency of owner-directed gazing in golden retrievers (gold) and Labrador retrievers divided by type. * = p < 0.05. Error bars show ± 1 SE.
